Create the supplyThe agentic build environmentWho builds the apps

Forge

Builds the supply as fast as agents can run.

StatusBuild
FIG · An agent assembles an app from units on the frame grid.

Forge is the agentic build environment for constructing and maintaining apps on the framework it is given: DAF today, Frame in the DAAC-native form, or another substrate once Forge holds its contracts and source. It is the operational form of the shift behind the whole suite: humans build the agents, and the agents build the apps.

Agents work in four modes. Compose assembles an app from the substrate’s existing units; Extend proposes a new reusable unit when the substrate does not yet cover a need; Convert turns a running system into an app on a target substrate, data migration included; Maintain carries substrate evolution down into apps already shipped. Human supervision sits at the scope boundaries (which apps, which spec changes, which extensions are admitted), not on every line.

Forge and Frame are separable in both directions: a customer can run Frame with human developers and no Forge, and Forge can be tasked against DAF or another framework. Frame stays its reference substrate, the one whose component doctrine gives the agents their richest starting pattern; a Datastake-style deployment uses both, which is how the ecosystem can add apps in hard sectors faster than headcount ever allowed.

Where it sits. Framework-agnostic: builds against DAF, Frame or another substrate it holds contracts for, with Frame the DAAC-native reference. Independent of Spine and Scribe; neither Frame nor Forge requires the other.

01What it does
01
Compose · Extend · Convert · Maintain
Agents assemble apps from the substrate’s own units, propose new ones when reality outruns it, convert running systems, and keep shipped apps current.
02
Supervised at the seams
Humans approve scope changes and substrate additions; agents do the construction in between.
03
Output that does not scale with hiring
Output scales with agent capability, not hiring, so new apps in new sectors stop being a staffing question.
02How it works
FIG · Humans build the agents; the agents build the apps
  1. 01Compose
    Higher volume

    Agents select and parameterise the target substrate’s existing units to assemble an app against its spec.

  2. 02Extend
    Higher stakes

    When the substrate can’t cover a need, agents propose a new reusable unit, admitted under that substrate’s own rules.

  3. 03Convert
    From a running system

    An existing system becomes an app on a target substrate, behaviour parity and data fidelity verified against the source.

  4. 04Maintain
    Nothing rots

    Substrate evolution is carried down into apps already shipped, so no app falls silently behind substrate.

Substrate reconnaissance first, fail-closed; new units pass a curation-review admission gate. Supervised at the seams, scope changes and unit admission, not every config line.

03How it connects
FIG · Forge is framework-agnostic; Frame is its DAAC-native reference substrate, not a dependency, and neither requires the other
Feeds inHands onConsoleobservesGimbalproposes allocationMasonsibling builderreference substratesuperviseFrametoolHumansexternalships →appscapture the fieldForgeThe agentic build environmentWho builds the apps
Forge, the subject hereCopper edges, what it hands on to the marketDashed leads, tooling that watches, never steers
04What it’s for
FIG · New supply, at agent pace
ForgeApps in hard sectorsfirst-party supplyOutput that does not scale with hiringbounded by agent capabilityIntelligencethe demand-side payoffpriced access →· Risk management· Impact measurement· Due diligence· Opportunity assessment
05What it reports
Indicative telemetry
The method runs supervised today; the packaged agent is still to ship
Born wired · awaiting first run
Reconnaissance hit-rateindicative

how often a capability was reused rather than rebuilt, the flagship metric

Work by modeindicative

composed, extended, converted and maintained, per build

Human checkpointsindicative

supervision events at the seams

Apps behind substrateindicative

drift of shipped apps against the evolving framework each was built on

Builds · apps shippedindicative

the lifecycle counters

Born wired, the telemetry shape is fixed before the packaged agent runs a build. The method already works under supervision: a production app has been rebuilt on a new substrate and verified to behave identically. Until the agent itself runs, the snapshot reads “connected, awaiting builds”, every field zeroed. Nothing is faked.

06Where it does the work